Prescription and self-reported use of endocrine therapy among Black hormone receptor-positive breast cancer survivors in the Detroit Research on Cancer Survivors cohort

Abstract

Purpose

Black women with hormone receptor-positive (HR +) breast cancer are twice as likely as White women to have weakly HR + tumors (1–10% positive cells). Patients with weakly HR + tumors are less frequently prescribed ET and have 60% higher mortality than strongly HR + tumors (> 10% positive cells). We evaluated factors associated with ET prescription and self-reported use among Black women with HR + breast cancer.

Methods

Among 922 Detroit ROCS participants, we evaluated associations between demographics, socioeconomic status, and health, tumor, oncologist, and hospital characteristics and ET prescription intent and self-reported ET use. Logistic mixed-effects regression was used to account for oncologist and hospital group effects.

Results

Oncologists intended to prescribe ET to 83.4% of participants (n = 769), of which 54.4% (n = 502) reported use. In multivariable models, participants with weakly HR + tumors were 90% less likely to be prescribed ET (OR = 0.10, p < 0.0001). Other significant characteristics of ET prescription included a BMI of 25–29.9 kg/m² (OR = 0.45, p = 0.0085), HR positivity > 90% vs. 11–90% (OR = 0.37, p = 0.00045), unknown HR percentage (OR = 0.12, p < 0.0001), OncotypeDx testing (OR = 2.65, p < 0.0001), and receiving radiation (OR = 2.20, p = 0.00016). Self-reported ET use was lower among those with lower health literacy (OR = 0.017, p < 0.001), weak HR positivity (OR = 0.46, p = 0.0053), unknown HR percentage (OR = 0.074, p = 0.034), and older age at diagnosis (OR = 0.88, p = 0.002). Increased ET use was associated with an income between $60,000-$79,900 vs. < $20,000 (OR = 1.54, p = 0.035), higher comorbidity count (OR = 1.09, p = 0.0054), distant stage (OR = 2.03, p = 0.029), and surgery (OR = 2.35, p = 0.001).

Conclusion

Identifying multilevel factors related to ET use may inform strategies to improve ET uptake and survival among Black women with HR + breast cancer.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10549-025-07816-1.

Introduction

Black women in the United States are 50% more likely to die from hormone receptor-positive (HR +), HER2-negative breast cancer than White women, even when accounting for factors including tumor stage and other clinical characteristics [1–4]. While lack of access to high quality healthcare is certainly a driving factor in this disparity, there is also evidence for underlying racial differences in tumor biology among HR +/HER2-negative breast tumors [5, 6]. We have shown that Black women are about twice as likely to have HR +/HER2-negative tumors with lower levels of HR staining (weakly HR + : 1–10% of cells staining positive for estrogen receptor and/or progesterone receptor) compared to White women, and women with weakly HR + tumors have approximately two-fold higher breast cancer mortality compared to women with strongly HR + tumors [7]. Importantly, we and others have shown that women with weakly HR + tumors who are prescribed endocrine therapy (ET) have substantially improved survival compared to those that were not prescribed ET [7, 8], yet these women were also ~ 20–30% less likely to be prescribed ET by their physicians [7, 8].

Breast cancer management requires patients and physicians to navigate a complex array of treatment options including ET, which is pivotal for those diagnosed with HR + breast cancer particularly for preventing breast cancer recurrence [9]. Clinical guidelines universally recommend ET when HR expression is detected in at least 1% of tumor cells. However, the decision becomes more nuanced for women with weakly HR + tumors as these patients might benefit from additional discussions on the comparative risks of ET and potential benefit of chemotherapy. In practice, treatment choices for weakly HR + cancers seem to align more closely with HR-negative protocols, excluding ET in favor of chemotherapy, which points to a gap between guideline recommendations and clinical practice [10]. This is particularly impactful for Black women who are most likely to have weakly HR + tumors and also underlines the need to address specific biological and systemic factors affecting treatment for all Black women with HR + breast cancer.

Adherence to ET is another critical factor influencing outcomes [11–13]. Adherence to ET can be difficult to assess at both the clinical and research levels but has been shown to be troublingly low, with a recent systematic review estimating 5-year adherence at 66% [14]. Adherence rates may be especially low among Black women and those from lower socioeconomic backgrounds, likely exacerbating disparities in breast cancer outcomes in these populations [15–17]. The reasons for discontinuing ET are varied and likely population- and context-specific, ranging from its prolonged use to adverse effects like hot flashes, fatigue, and more severe complications such as osteoporosis [18, 19]. The landscape of treatment decisions from the perspectives of both prescription and adherence is further complicated by disparities in healthcare access and quality. Black women, for instance, are less likely to receive optimal cancer treatments including surgical interventions and adjuvant therapies [20–22]. While factors such as limited health literacy and less access to experienced physicians and quality medical resources [21, 23–27] have been associated with suboptimal cancer care, a better understanding of the barriers to receiving and adhering to ET among Black women is critical so that we can identify areas for culturally appropriate intervention to improve the equity of care of an underserved patient population.

The goal of this analysis was to explore and identify multilevel factors—including demographic, socioeconomic, and health factors at the individual level, tumor and clinical characteristics, and physician and treating hospital characteristics—associated with prescription and self-reported use of ET among Black women with invasive HR +/HER2- breast cancer. We utilized survey data from the Detroit Research on Cancer Survivors cohort, one of the largest studies conducted exclusively among Black cancer survivors in metropolitan Detroit—linked with the Metropolitan Detroit Cancer Surveillance System registry, the National Plan and Provider Enumeration System (NPPES) and the American Hospital Association (AHA)—to model these relationships, hypothesizing that unique sets of multilevel associations with prescription and self-reported use of ET in this population.

Materials and methods

Study population

We included data from female participants with invasive breast cancer enrolled in the Detroit Research on Cancer Survivors (ROCS) cohort study of Black cancer survivors, which has been described previously [28, 29]. Eligible Detroit ROCS cases between the ages of 20 and 79 were identified through the Metropolitan Detroit Cancer Surveillance System (MDCSS), and included first primary invasive female breast, colorectal, lung, and prostate cancers diagnosed on or after January 1, 2013, and endometrial and other young onset (age at diagnosis ≥ 20 and ≤ 49) cancer cases diagnosed on or after January 1, 2016. First primary status was determined by the Surveillance, Epidemiology, and End Results (SEER) Program’s Multiple Primary Coding Rules [30]. Individuals with prior cancer diagnoses, in situ disease, or bilateral disease were not excluded. All participants self-identified as Black or African American. A total of 5069 survivors were enrolled into Detroit ROCS, which included completion of a baseline questionnaire with annual follow-up to ascertain and update medical history, health-related quality of life, health behaviors, and treatment information. Participants complete questionnaires either online, over the phone with a trained interviewer, or by returning a written copy. For this analysis, we included all hormone receptor-positive (HR + : estrogen receptor [ER] and/or progesterone receptor [PR] positive) breast cancer ROCS cases for whom percent ER and PR could be abstracted from pathology reports, resulting in a final sample size of 922 women. The eligibility criteria and selection process for the analytic cohort are summarized in Fig. 1. The Wayne State University Institutional Review Board approved this study (WSU IRB#050417M1F), and informed consent was obtained from all subjects prior to participation.

Fig. 1.

Flow diagram summarizing the research design. MDCSS Metropolitan Detroit Cancer Surveillance System

Patient and tumor-level variables and covariate data

Participants self-reported demographic, socioeconomic, and health status variables at enrollment, including education, marital status, annual household income, health insurance status at diagnosis, health literacy confidence, comorbidities, and smoking history prior to diagnosis on enrollment questionnaires. Body mass index (BMI; body weight in kg/height in meters squared) was calculated from self-reported height and weight one year before cancer diagnosis. The percentage of cells staining positive for ER and PR was abstracted from pathology reports. Using the maximum percent staining value among the two hormone receptors, we categorized tumors as weakly HR + (1–10%), moderately HR + (11–90%), and strongly HR + (> 90%). Clinical characteristics including HER2 status, stage at diagnosis using the Surveillance Epidemiology and End Results (SEER) summary staging system, OncotypeDx testing status and score, and first course of treatment (defined as receipt of therapy for a cancer diagnosis before disease progression or recurrence including surgery), adjuvant chemotherapy, endocrine therapy, and radiation therapy were obtained via MDCSS registry linkage. Surgery, chemotherapy, and radiation therapy variables used standard SEER coding conventions and were further categorized into binary “yes/no” variables based on indications of whether the patient was scheduled or prescribed that therapy.

Hospital and physician variables

Primary treating physicians were identified by National Provider Identifier (NPI) using MDCSS data and validated by reviewing source documents in the electronic registry database (SEER*DMS). When more than one physician across multiple specialties were identified, medical oncologist was selected, and if no medical oncologist was listed, then surgical oncologist was selected. Physician data were obtained from the National Plan and Provider Enumeration System (NPPES) Monthly Downloadable File, downloaded from the Centers for Medicare & Medicaid Services website (cms.gov) on September 26, 2023. Physician variables included physician credentials, gender, specialty taxonomy, whether they were the sole proprietor of their practice, and the number of years in practice. Primary treating hospital was coded using a unique identifier and linked to data from the American Hospital Association (AHA) to obtain characteristics including hospital size (number of inpatient beds), presence of a resident training program, and profit vs. non-profit status.

Endocrine therapy variables

ET data were obtained from two sources—MDCSS and self-report on participant questionnaires—to derive two separate variables corresponding to physician intent to prescribe ET and patient-reported use of ET. MDCSS ET data were abstracted from medical records and coded by trained abstractors using North American Association of Central Cancer Registry (NAACR) standard procedures. This includes information from both structured data fields and provider documentation. These provider notes are not part of an open notes system but are derived from the clinical documentation available in the MDCSS medical records. MDCSS treatment data capture only first course of therapy as recorded at the time of initial treatment. Therefore, it does not account for subsequent treatment decisions, changes in prescribing, or therapies initiated later in the patient’s care. MDCSS-derived ET data were operationalized to reflect physician intent to prescribe, categorized as “No” only when there was no indication of physician recommendation for prescription. Intent to prescribe ET was defined as “Yes” when physicians prescribed ET or recommended ET but patients did not take it due to refusal or contraindications. Patient-reported use of ET was obtained from enrollment and yearly follow-up questionnaires using responses to the questions “Have you received hormone therapy (in any form) to treat your breast cancer” (enrollment survey), “In the past year have you received hormone therapy (in any form) to treat your breast cancer?” (Year 1 follow-up survey), or “Since the last survey, have you received hormone therapy (in any form) to treat your breast cancer?” (follow-up surveys in Year 2 and beyond). Enrollment survey ET data were supplemented by reviewing the reported current prescription medication list, where responses corresponded to known ET medications.

Statistical analysis

Statistical analyses were carried out using R software v4.3.2 (https://cran.r-project.org/). A nominal two-sided alpha value of 0.05 was set to determine statistical significance. Physician intent to prescribe endocrine therapy (ET) was categorized as a binary variable based on data from the first course of treatment. Self-reported use of ET was categorized as a binary variable at each time point for which data were available in enrollment or follow-up questionnaires.

Associations between study factors and outcomes were assessed using logistic regression models. Statistical significance for individual independent variables in logistic regression models were assessed based on the Walds test statistic. For univariate analyses (Tables 1, 2, 4, and 5), univariable logistic regression was used to evaluate the relationship between individual patient, physician, or hospital factors and the outcomes of interest. Associations between physician intent to prescribe ET and individual (1) patient and tumor-level characteristics were assessed using simple logistic regression with the “glm” function and (2) using logistic mixed-effects regression with the “glmer” function for physician and hospital characteristics, accounting for clustering within physician and hospital, respectively.

Table 1.

Univariate associations between patient & clinical characteristics and physician intent to prescribe endocrine therapy

	Did not intend to prescribe ET (n = 153)		Intended to prescribe ET (n = 769)
	n	%	n	%	OR	95% CI	p-value
Demographic
Age at diagnosis
< 50	43	16.9	211	83.1	1.00 (ref)	–	–
50–59	48	17.6	224	82.4	0.95	0.60–1.50	0.83
60–69	45	15.8	239	84.2	1.08	0.69–1.71	0.73
70–79	17	15.1	95	84.8	1.14	0.62–2.10	0.68
Socioeconomic
Education
Less than HS	8	11.4	62	88.6	1.00 (ref)	–	–
HS	36	18.2	162	81.8	0.58	(0.26–1.32)	0.19
Some college	75	15.7	403	84.3	0.69	(0.32–1.50)	0.36
Professional/graduate	33	20.2	130	79.8	0.51	(0.22–1.17)	0.11
Insurance
Medicaid ± other	50	19.7	204	80.3	0.77	(0.50–1.19)	0.23
Medicare ± other	51	14.8	294	85.2	1.08	(0.71–1.65)	0.72
Other	2	28.6	5	71.4	0.47	(0.09–2.49)	0.37
Unknown	1	20	4	80	0.75	(0.08–6.85)	0.8
Private	49	15.8	261	84.2	1.00 (ref)	–	–
Income
< $20 k	53	15.9	280	84.1	1.00 (ref)	–	–
$20 k–39.9 k	34	16.6	171	83.4	0.95	(0.59–1.52)	0.84
$40 k–59.9 k	21	15.2	117	84.8	1.05	(0.61–1.82)	0.85
$60 k–79.9	14	18.4	62	81.6	0.84	(0.44–1.61)	0.6
$80 k +	17	17.3	81	82.7	0.9	(0.49–1.64)	0.74
Health literacy
Extremely confident	85	16.4	434	83.6	1.00 (ref)	–	–
Quite a bit	32	15.8	170	84.2	1.04	(0.67–1.62)	0.86
Somewhat	25	20.2	99	79.8	0.78	(0.47–1.28)	0.32
A little bit	7	17.5	33	82.5	0.92	(0.39–2.15)	0.85
Not at all	3	11.1	24	88.9	1.57	(0.46–5.33)	0.47
Health status
Total # comorbidities
0	15	23.1	50	76.9	1.00 (ref)	–	–
1–3	26	15.6	141	84.4	1.54	(0.83–2.87)	0.17
4 +	55	15.8	294	84.2	1.6	(0.84–3.05)	0.15
Cardiovascular comorbidities
0–1	131	16.2	680	83.8	1.00 (ref)	–	–
2 +	22	19.8	89	80.2	1.28	(0.77–2.12)	0.33
BMI
< 25 kg/m2	16	13	107	87	1.00 (ref)	–	–
25–29.9 kg/m2	57	24.9	172	75.1	0.45	(0.25–0.82)	0.01
30–39.9 kg/m2	57	14.1	348	85.9	0.91	(0.50–1.65)	0.76
40 + kg/m2	19	13.1	126	86.9	0.99	(0.49–2.02)	0.98
BMI (per kg/m2)	Median	IQR	Median	IQR
	31	8.3	31.4	9.6	1	(0.98–1.02)	0.93
Smoking	n	%	n	%
Never	123	17	600	83	1.00 (ref)	–	–
Ever	29	14.9	165	85.1	1.17	(0.30–4.49)	0.49
Tumor characteristics
ER/PR percent staining
1–10%	21	42.9	28	57.1	0.09	(0.05–0.18)	< 0.0001
11–90%	29	15.4	159	84.6	0.37	(0.22–0.63)	0.00032
> 90%	31	6.4	456	93.6	1.00 (ref)	–	–
Unknown	72	36.4	126	63.6	0.12	(0.08–0.19)	< 0.0001
HER2 status
Negative	125	17	610	83	1.00 (ref)	–	–
Positive	25	16	131	84	1.07	(0.67–1.71)	0.77
Unknown	3	9.7	28	90.3	1.91	(0.57–6.38)	0.29
Stage
Localized	103	18.4	457	81.6	1.00 (ref)	–-	–-
Regional	42	13	280	87	1.5	(1.02–2.21)	0.04
Distant	8	20	32	80	0.9	(0.4–2.01)	0.8
OncotypeDx Score
Low risk	17	8.6	101	91.4	1.00 (ref)	–	–
Intermediate risk	13	11.4	36	88.6	0.73	(0.34–1.56)	0.43
High risk	2	2.3	451	94.7	1.7	(0.38–7.68)	0.49
Not performed	121	21.2	451	78.8	0.35	(0.20–0.6)	0.0014
Treatment
Surgery
Yes	135	15.9	713	84.1	1.76	(0.99–3.14)	0.056
No	17	25	51	75	1.00 (ref)	–	–
Chemotherapy
Yes	82	18	374	82	0.83	(0.59–1.18)	0.3
No	71	15.5	388	84.5	1.00 (ref)	–	–
Radiation
Yes	90	13.8	561	86.2	1.94	(1.36–2.79)	0.00029
No	63	23.8	202	76.2	1.00 (ref)	–	–

Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated using univariable logistic regression. P values are two-sided and nominal

Table 2.

Univariate associations between physician & hospital characteristics and physician intent to prescribe endocrine therapy

	Did not intend to prescribe ET (n = 153)		Intended to prescribe ET (n = 769)
	n	%	n	%	OR	95% CI	P-value
Physician characteristics
Credential
DO	9	18.0	41	82.0	1.09	(0.31–3.85)	0.9
Unknown	1	16.7	5	83.3	1.46	(0.06–38.7)	0.8
MD	143	17.7	663	82.3	1.00 (ref)	–	–
MD + other	0	0.0	60	100.0	NA	NA	NA
Gender
Female	84	18.5	369	81.5	1.00 (ref)	–	–
Male	68	16.7	395	83.3	1.73	(0.07–40.7)	0.36
Unknown	1	14.7	5	85.3	1.44	(0.66–3.12)	0.74
Specialty
Internal medicine	7	16.7	35	83.3	0.39	(0.09–1.56)	0.18
Med-onc/Heme-onc	53	8.4	558	91.6	1.00 (ref)	–	–
Other	3	15.8	16	84.2	0.60	(0.12–3.01)	0.54
Surgery	89	36.0	158	64	0.12	(0.06–0.24)	< 0.0001
Sole proprietor
Yes	28	34.1	54	65.9	0.49	(0.20–1.39)	0.19
No	124	14.7	708	85.3	1.00 (ref)	–	–
Unknown	1	9.1	10	90.9	2.93	(0.17–43.1)	0.43
Years in practice
	Median	IQR	Median	IQR
	10.5	3.0	9.1	3.1	0.74	(0.63–0.89)	0.000349
Hospital characteristics
Bed size	n	%	n	%
50–99	21	8.1	239	91.9	2.55	(0.34–19.1)	0.36
100–499	45	18.1	204	81.9	1.00 (ref)	–	–
500 +	87	21.3	321	78.7	1.29	(0.46–3.7)	0.63
Resident training
No	4	6.3	59	93.7	1.00 (ref)	–	–
Yes	149	17.4	707	82.6	0.28	(0.04–2.18)	0.22
Class
For profit	26	33.3	52	66.7	1.00 (ref)	–	–
Non-profit	127	15.1	714	84.9	2.39	(0.75–8.26)	0.14

Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated using univariable logistic regression. P values are two-sided and nominal

Table 4.

Univariate associations between patient & clinical characteristics and self-reported use of endocrine therapy

	Never reported ET use (n = 420)		Reported ET use (n = 502)
	n	%	n	%	OR	95% CI	p-value
Demographic
Age at diagnosis
< 50	100	39.4	154	60.6	1.00 (ref)	–	–
50–59	129	47.4	143	52.6	0.72	0.51–1.02	0.063
60–69	136	47.9	148	52.1	0.71	0.50–0.99	0.047
70–79	55	49.6	57	50.4	0.66	0.42–1.03	0.069
Socioeconomic
Education
Less than HS	42	60	28	40	1.00 (ref)	–	–
HS	105	53	93	47	1.07	(0.67–1.71)	0.76
Some college	206	43.2	272	56.8	1.56	(1.02–2.39)	0.042
Professional/graduate	62	38	101	62	1.63	(1.02–2.60)	0.04
Insurance
Medicaid ± other	120	47.2	134	52.8	0.87	(0.67–1.14)	0.29
Medicare ± other	161	46.8	184	53.2	0.8	(0.62–1.02)	0.081
Other	6	85.7	1	14.3	0.25	(0.05–1.36)	0.11
Unknown	129	41.6	181	58.4	1.00 (ref)	–	–
Private	3	60	2	40	0.38	(0.06–2.23)	0.29
Income
< $20 k	170	51.2	163	48.8	1.00 (ref)	–	–
$20 k–39.9 k	92	44.9	113	55.1	1.17	(0.88–1.56)	0.27
$40 k–59.9 k	54	39.1	84	60.9	1.16	(0.84–1.60)	0.35
$60 k–79.9	28	36.8	48	63.2	1.71	(1.16–2.53)	0.0072
$80 k +	35	35.7	63	64.3	1.64	(1.16–2.32)	0.0057
Health literacy
Extremely confident	208	40.2	311	59.8	1.00 (ref)	–-	–-
Quite a bit	91	45	111	55	0.76	(0.58–0.99)	0.04
Somewhat	71	57.3	53	42.7	0.61	(0.43–0.86)	0.0043
A little bit	25	62.5	15	37.5	0.56	(0.31–1.01)	0.05
Not at all	22	81.5	5	18.5	0.24	(0.10–0.57)	0.0013
Health status
Total # comorbidities
0	31	47.7	34	52.3	1.00 (ref)	–	–
1–3	231	45.8	273	54.2	1.35	(0.88–2.08)	0.17
4 +	157	45.1	192	54.9	1.44	(0.93–2.24)	0.1
Cardiovascular comorbidities
0–1	55	50	56	50	1.00 (ref)	–
2 +	365	45	446	55	0.91	(0.79–1.05)	0.19
BMI
< 25 kg/m2	60	48.8	63	51.2	1.00 (ref)	–	–
25–29.9 kg/m2	109	47.8	120	52.2	1.09	(0.76–1.56)	0.64
30–39.9 kg/m2	176	43.5	229	56.5	1.18	(0.85–1.64)	0.33
40 + kg/m2	63	43.4	82	56.6	1.42	(0.96–2.09)	0.076
BMI (per kg/m2)	Median	IQR	Median	IQR
	30.9	9.3	31.7	9.4	1.03	(1.02–1.04)	0.021
Smoking	n	%	n	%
Never	322	44.6	401	55.4	1.00 (ref)	–	–
Ever	97	50	97	50	0.76	(0.58–0.99)	0.043
Tumor characteristics
ER/PR percent staining
1–10%	31	63.3	18	36.7	0.47	(0.28–0.80)	0.0051
11–90%	85	45.5	103	54.5	0.84	(0.64–1.10)	0.22
> 90%	205	42.1	282	57.9	1.00 (ref)	–	–
Unknown	99	50	99	50	0.71	(0.54–0.93)	0.013
HER2 status
Negative	337	45.9	398	54.1	1.00 (ref)	–	–
Positive	67	42.9	89	57.1	1.17	(0.89–1.55)	0.26
Unknown	16	51.6	15	48.4	0.84	(0.46–1.52)	0.55
Stage
Localized	256	45.8	304	54.2	1.00 (ref)	–	–
Regional	146	45.3	176	54.7	1.09	(0.87–1.36)	0.46
Distant	18	45	22	55	1.5	(0.86–2.62)	0.15
OncotypeDx Score
Low risk	79	40.1	118	59.9	1.00 (ref)	–	–
Intermediate risk	51	44.7	63	55.3	1	(0.69–1.44)	0.99
High risk	18	47.4	20	52.6	0.91	(0.52–1.59)	0.74
Not performed	271	47.5	301	52.5	0.8	(0.62–1.04)	0.092
Treatment
Surgery
Yes	43	63.2	25	36.8	1.89	(1.18–3.02)	0.0074
No	371	43.8	477	56.2	1.00 (ref)	–	–
Chemotherapy
Yes	213	46.1	250	53.9	1.07	(0.87–1.32)	0.53
No	207	45.1	252	54.9	1.00 (ref)	–	–
Radiation
Yes	135	51.1	130	48.9	1.31	(1.03–1.66)	0.026
No	280	43	371	57	1.00 (ref)	–	–

Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated using univariable logistic regression. P values are two-sided and nominal

Table 5.

Univariate associations between physician & hospital characteristics and self-reported use of endocrine therapy

	Never reported ET use (n = 420)		Reported ET use (n = 502)
	n	%	n	%	OR	95% CI	P-value
Physician characteristics
Credential
DO	28	56	22	44	0.67	(0.41–1.08)	0.1
Unknown	1	16.7	5	83.3	1.56	(0.47–5.16)	0.46
MD	365	45.2	442	54.8	1.00 (ref)	–	–
MD + other	27	45	33	55	1.19	(0.79–1.80)	0.41
Gender
Female	204	45	249	55	1.00 (ref)	–	–
Male	216	46.6	248	53.4	1.54	(1.25–1.90)	0.66
Unknown	1	16.7	5	83.3	0.95	(0.29–3.16)	0.48
Specialty
Internal medicine	18	42.9	24	57.1	1.28	(0.78–2.10)	0.34
Med-onc/Heme-onc	269	44.2	340	55.8	1.00 (ref)	–	–
Other	11	57.9	8	42.1	0.57	(0.26–1.26)	0.16
Surgery	122	49.4	125	50.6	0.85	(0.67–1.08)	0.19
Sole proprietor
Yes	379	45.7	451	54.3	1.00 (ref)	–	–
No	1	50	10	50	0.97	(0.67–1.40)	0.88
Unknown	41	9.1	41	90.9	2.26	(0.94–5.46)	0.07
Years in practice
	Median	IQR	Median	IQR
	9.3	3.2	9.5	3	1.02	(0.97–1.07)	0.37
Hospital characteristics
Bed size	n	%	n	%
50–99	132	50.8	128	49.2	0.89	(0.69–1.15)	0.16
100–499	114	45.8	135	54.2	0.83	(0.64–1.07)	0.36
500 +	174	42.5	235	57.5	1.00 (ref)	–	–
Resident training
No	21	33.3	42	66.7	1.00 (ref)	–	–
Yes	400	46.7	457	53.3	0.77	(0.51–1.15)	0.2
Class
For profit	37	47.4	41	52.6	1.00 (ref)	–	–
Non-profit	384	45.6	458	54.4	1.04	(0.72–1.50)	0.82

Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated using univariable logistic regression. P values are two-sided and nominal

All models assessing self-reported use of ET employed multilevel logistic regression with the “glmer” function, with a random-effects clustering term based on unique patient identifiers to allow for multiple ET use measurements over time, and with additional random-effects clustering terms based on physician or hospital identifier as needed. For multivariable analyses (Tables 3 and 6), models were built using backward selection with the “GLMERSelect” function, where our full model included all variables individually associated with the outcome at P ≤ 0.10. Results from all models are reported as odds ratios (ORs) with 95% confidence intervals and nominal two-sided P values.

Table 3.

Multivariable model assessing physician intent to prescribe endocrine therapy

Variable	OR	95% CI	P value
BMI (25–29.9 kg/m2 vs. < 25 kg/m2)	0.41	(0.21–0.80)	0.0085
BMI (30–39.9 kg/m2 vs. < 25 kg/m2)	0.81	(0.43–1.54)	0.53
BMI (40 + kg/m2 vs. < 25 kg/m2)	0.81	(0.37–1.75)	0.59
BMI (Unknown vs. < 25 kg/m2)	0.54	(0.15–2.01)	0.36
ER/PR percent staining (1–10% vs. > 90%)	0.1	(0.05–0.20)	< 0.0001
ER/PR percent staining (11–90% vs. > 90%)	0.37	(0.21–0.65)	0.00045
ER/PR percent staining (Unknown vs. > 90%)	0.12	(0.07–0.19)	< 0.0001
Regional vs Local	1.79	(1.16–2.77)	0.0085
Distant vs Local	1.17	(0.48–2.85)	0.72
OncotypeDx (Tested vs. untested)	2.65	(1.66–4.22)	< 0.0001
Radiation (Yes vs. no)	2.2	(1.46–3.31)	0.00016

Multivariable logistic regression was used to estimate ORs and 95% CIs. P values are two-sided and nominal

Table 6.

Multivariable model assessing self-reported use of endocrine therapy

Variable	OR	95% CI	P value
Age at diagnosis (years)	0.88	(0.81–0.95)	0.002
Health literacy (Quite a bit vs. extremely confident)	0.74	(0.56–0.98)	0.038
Health literacy (Somewhat vs. extremely confident)	0.61	(0.42–0.87)	0.0067
Health literacy (A little bit vs. extremely confident)	0.64	(0.35–1.18)	0.16
Health literacy (Not at all vs. extremely confident)	0.17	(0.06–0.48)	0.0008
Income ($20 k–39.9 k vs. < $20 k)	1.1	(0.82–1.47)	0.53
Income ($40 k–59.9 k vs. < $20 k)	0.98	(0.71–1.37)	0.92
Income ($60 k–79.9 vs. < $20 k)	1.54	(1.03–2.31)	0.035
Income ($80 k + vs. < $20 k)	1.4	(0.97–2.03)	0.074
Number of comorbidities	1.09	(1.03–1.17)	0.0054
Smoking (Ever vs. never)	0.76	(0.57–1.02)	0.068
Stage 2 vs 1	1.12	(0.89–1.41)	0.33
Stage 3 vs 1	2.03	(1.08–3.83)	0.029
ER/PR percent staining (1–10% vs. > 90%)	0.46	(0.27–0.80)	0.0053
ER/PR percent staining (11–90% vs. > 90%)	0.82	(0.62–1.08)	0.16
ER/PR percent staining (Unknown vs. > 90%)	0.74	(0.55–0.98)	0.034
Surgery (Yes vs. no)	2.35	(1.38–3.99)	0.0015

Multivariable logistic regression was used to estimate ORs and 95% CIs. P values are two-sided and nominal

Results

Physician intent to prescribe endocrine therapy

Our final analytic sample included 922 Detroit ROCS HR + participants with invasive breast cancer. Median follow-up time was 3 years (range 1–6 years). Among all cases, the majority (83.4%) were recommended to receive ET by their physicians, while just over half (54.4%) reported using ET across the follow-up period. Table 1 details univariable associations between demographic and health status characteristics regarding physician intent to prescribe ET. Patients classified as overweight (BMI 25–29.9 kg/m²) were significantly less likely to be recommended or prescribed ET than those under 25 kg/m² (OR = 0.45, 95% CI 0.25–0.82), while no associations were observed for patients with BMI ≥ 30 kg/m². With respect to clinical and treatment variables, patients with tumors showing 1–10% and 11–90% HR staining were 91% (OR = 0.09, 95% CI 0.05–0.18) and 63% (OR = 0.37, 95% CI 0.22–0.63) less likely to be recommended or prescribed ET, respectively, compared to those with > 90% HR staining, while those with unknown HR staining percentages also had a reduced likelihood (OR = 0.12, 95% CI 0.08–0.19). Women with regionally staged tumors had a 1.5-times (95% CI 1.02–2.21) higher likelihood of ET recommendation or prescription compared to women with localized tumors. Interestingly, patients who did not undergo OncotypeDx scoring were less likely to be recommended or prescribed ET compared to those with a low risk OncotypeDx score (OR = 0.35, 95% CI 0.20–0.60). Patients undergoing radiation (OR = 1.94, 95% CI 1.36–2.79) and surgery (OR = 1.76, 95% CI 0.99–3.14) were more likely to be recommended or prescribed ET, although surgery was only marginally statistically significant. Among physician and hospital characteristics (Table 2), only specialty (Surgery vs. Medical oncology: OR = 0.12, 95% CI 0.06–0.24) and years in practice (OR = 0.74, 95% CI 0.63–0.89) were associated with ET.

In a multivariable model of physician intent to prescribe ET (Table 3), only patient- and tumor-level characteristics remained associated with the outcome. Patients classified as overweight (BMI 25–29.9 kg/m²) were significantly less likely to be prescribed ET than those under 25 kg/m² (OR = 0.41, 95% CI 0.21–0.80). Patients with tumors characterized as 1–10% (OR = 0.10, 95% CI 0.05–0.20), 11–90% (OR = 0.37, 95% CI 0.21–0.65), and unknown (OR = 0.12, 95% CI 0.07–0.19) HR staining were significantly less likely to receive ET compared to those with greater than 90% HR staining. Additionally, those diagnosed with regional breast cancer were 1.79 times (95% CI 1.16–2.77) more likely to be prescribed ET than those with locally staged tumors. Women who had OncotypeDx testing (OR = 2.65, 95% CI 1.66–4.22) and who were treated with radiation (OR = 2.20, 95% CI 1.46–3.31) were also more likely to be recommended or prescribed ET.

Self-reported use of endocrine therapy

We next examined individual relationships between individual-, physician-, and hospital-level variables and the self-reported use of ET among all participants (Tables 4, 5, 6) and among the subset whose physicians intended to prescribe ET at the time of first treatment (Tables S1–S3). Among all participants (Table 4), we observed that individuals aged over 50 were less inclined to report the use of ET compared to those under 50 years old (50–59 years: OR = 0.72, 95% CI 0.51–1.02; 60–69 years: OR = 0.71, 95% CI 0.50–0.99; 70–79 years: OR = 0.66, 95% CI 0.42–1.03). Moreover, individuals with some college education (OR = 1.56, 95% CI 1.02–2.39) or a professional/graduate-level education (OR = 1.63, 95% CI 1.02–2.60) were more likely to report ET usage than those with less than a high school education. Similarly, patients with annual incomes exceeding $60,000 reported higher ET usage compared to those with less than $20,000 annual income ($60–79.9 k: OR = 1.71, 95% CI 1.16–2.53; $80 k + : OR = 1.64, 95% CI 1.16–2.32). Notably, reports of ET usage significantly decreased as self-reported health literacy decreased. Specifically, patients who were not at all confident were 76% less likely to report ET usage compared to those who were extremely confident (OR = 0.24, 95% CI 0.10–0.57). Results among the subset whose physicians intended to prescribe ET at the time of first treatment were largely similar (Table S1), except the magnitude and statistical significance of some associations differed. Age trends were similar but only statistically significant for the oldest compared to the youngest age group (70–79 years: OR = 0.74, 95% CI 0.52–1.07). Education and health literacy effects were consistent, while income effects were slightly attenuated for the $60–79.9 k (OR = 1.51, 95% CI 0.99–2.29) and $80 k + (OR = 1.29, 95% CI 0.89–1.88) groups.

Health status and clinical characteristics associated with reports of ET use were then explored. Among all participants, we found that body mass index (BMI) as a continuous variable showed a significant association with reports of ET use where a 1 kg/m² increase in BMI corresponded to a 3% (OR = 1.03, 95% CI 1.02–1.04) increase in likelihood of reporting ET use. Furthermore, ever smokers were less likely to report ET usage than never smokers (OR = 0.76, 95% CI 0.58–0.99). Women with tumors that were either weakly HR positive or had unknown HR percent were significantly less likely to report ET usage (weakly HR + : OR = 0.47, 95% CI 0.28–0.80; Unknown: OR = 0.71, 95% CI 0.54–0.93) than women with strongly HR + tumors. Conversely, women who had undergone surgery (OR = 1.89, 95% CI 1.18–3.02) or radiation therapy (OR = 1.31, 95% CI 1.03–1.66) were more likely to report ET usage than women who did not have those treatments. In the subset analysis (Table S1), BMI, tumor, and treatment effects were largely consistent with the overall analysis. Across credential type, gender, specialty, sole proprietorship status, hospital bed size, presence of residency training programs, and hospital class, no variables reached statistical significance in either the overall or the subset analyses (Table 5, Table S2).

Multivariable analyses were then conducted to evaluate associations with self-reported ET use among all participants (Table 5) and the subset whose physicians intended to prescribe ET at the time of first treatment (Table S3). Among all participants, as age of diagnosis increased, patients were less likely to report ET use (OR = 0.88, 95% CI 0.81–0.95). Consistent with our univariate analyses, self-reported health literacy was linked to differences in reported ET use, most notably for those who reported having no confidence at all (OR = 0.17, 95% CI 0.06–0.48) compared to those with high confidence levels. Patients who were"somewhat"or"quite a bit"confident were also less likely to report ET use compared to those who were extremely confident (OR = 0.61 and 0.74, respectively). Patients with an income between $60 k and $79.9 k were more likely to report ET use compared to those with an income less than $20 k (OR = 1.54, 95% CI 1.03–2.31). Although not statistically significant, patients with an income greater than $80 k were also more likely to report ET use (OR = 1.40, 95% CI 0.97–2.03). Increased reports of ET use were also associated with increasing number of comorbidities (OR = 1.09, 95% CI 1.03–1.17), distant compared to local cancer stage (OR = 2.03, 95% CI 1.08–3.83), and having surgery (OR = 2.35, 95% CI 1.38–3.99). HR staining once again showed importance as patients with tumors that stained 1–10% HR (OR = 0.46, p = 0.0053) and those with unknown HR percent staining (OR = 0.74, 95% CI 0.55–0.98) were significantly less likely to report ET use compared to patients with tumors staining greater than 90%.

In the multivariable model among the subset whose physicians intended to prescribe ET at first treatment (Table S3), age at diagnosis showed a similar inverse association with ET use as in Table 6 but did not reach statistical significance and income dropped out of the model entirely. Health literacy patterns were consistent, although the “quite a bit” confident group did not reach statistical significance (OR = 0.77, 95% CI 0.59–1.02). Number of comorbidities (OR = 1.08, 95% CI 1.01–1.15), stage 3 vs. stage 1 (OR = 2.94, 95% CI 1.52–5.69), and surgery (OR = 2.57, 95% CI 1.45–4.56) were significantly associated with higher ET use in both analyses. ER/PR staining showed weaker associations in Table S3 with 1–10% vs. > 90% (OR = 0.66, 95% CI 0.36–1.22, p = 0.187), 11–90% vs. > 90% (OR = 0.78, 95% CI 0.59–1.04, p = 0.090), and unknown vs. > 90% (OR = 0.76, 95% CI 0.56–1.05, p = 0.093) compared to the overall analyses.

Discussion

The goal of this analysis was to improve our understanding of the various factors influencing both prescription and self-reported use of ET among HR + breast cancer patients within a population-based cohort of Black breast cancer survivors. Our findings are novel, as we evaluated two related outcomes—physician intent to prescribe ET and patient adherence to ET—among Black women. Additionally, unlike previous studies that often focus exclusively on patient or physician and hospital-based variables, our study distinctly integrates both aspects. By analyzing demographic, clinical, and treatment-related variables alongside physician and hospital characteristics, we offer a comprehensive evaluation of potential determinants of ET prescription and usage.

We found that physicians were less likely to intend to prescribe ET to women with tumors exhibiting 1–10%, 11–90%, and unknown HR staining compared to those with over 90% HR staining. Patients with weakly HR + tumors or unknown HR staining percentages were also much less likely to report using ET compared to those with tumors that have over strongly HR + staining. This is consistent with previous research showing that clinicians are less inclined to prescribe ET for weakly positive breast cancer [7, 10], despite evidence-based guidelines recommending ET for all patients with ER + breast cancerl [18]. Clinicians may perceive weakly HR + breast cancer to be less sensitive to ET, contributing to their reluctance to prescribe ET for this group, particularly given potentially significant side effects such as hot flashes, fatigue, and bone density loss [18, 19]. This, however, is in contrast to national guidelines as well as findings from recent studies showing that women with lower levels of HR staining who receive ET have survival rates comparable to women with strongly HR + tumors [7, 8].

Other tumor characteristics were also associated with prescription and use of ET, namely OncotypeDx testing and tumor stage. Physicians use OncotypeDx testing to inform decisions about adjuvant chemotherapy treatment [31], but the test is not directly linked to adjuvant ET decision-making outside of the fact that recurrence scores inherently reflect estrogen signaling levels [32]. However, we found that undergoing OncotypeDx testing was associated with a substantial increase in the likelihood of physician intent to prescribe ET. This could be indicative of other factors related to medical eligibility for systemic therapy in general such as poor general health and fitness at the time of treatment, which could simultaneously disqualify a patient for chemotherapy and endocrine therapy [33]. While ET prescription rates were high overall regardless of stage (80–87%), we found a significant increase among patients diagnosed at regional compared to local stages. This may indicate that potential adverse side effects of ET are more heavily weighted in early-stage breast cancer cases, which are less likely to recur compared to tumors with lymph node involvement [34]. In contrast, however, women diagnosed with distant stage tumors were twice as likely to report using ET, possibly because ET can be a critical component of extending life expectancy and quality at this incurable stage [35].

We observed a non-linear association between BMI and physician intent to prescribe ET, where only overweight patients were significantly less likely to be prescribed ET compared to normal weight, which may be reflecting the fact that overweight and obesity have complicated relationships with breast cancer outcomes. These patients are at higher risk of developing cardiovascular disease and diabetes, and there is evidence for increased risks of developing these co-morbidities after using ET [36]. However, obesity is associated with worse disease-specific and overall survival, which would argue for more aggressive treatment including the use of ET [37]. There is also the issue that ET does not adequately suppress serum estrogen levels in obese patients, increasing risk of recurrence, although there is evidence that anti-estrogens are more effective than aromatase inhibitors [38–41].

Our findings also highlight a significant relationship between health literacy and self-reported use of ET, where individuals with low health literacy demonstrated markedly lower use of ET compared to their highly confident counterparts, particularly among those who were not at all confident. Health literacy has been shown to be lower in minoritized populations, which is related to issues of structural racism, socioeconomic status, resource availability, medical discrimination, poor communication and mistrust, and provider biases [42–44]. Previously studies have shown that health literacy is lower in minoritized populations, which can lead to a variety of health outcomes among breast cancer patients including developing chronic conditions, poor self-management and medication adherence, increased hospitalization, and worse quality of life [45–48].

This study has important limitations to consider. Our analysis relied on self-reported endocrine therapy (ET) use rather than insurance claims or pharmacy data, which were not available. Additionally, we were only able to evaluate physician intent to prescribe ET at the first course of treatment because of limited access to complete medical records and reliance on MDCSS registry data.

Although under-reporting of ET use is a recognized limitation of population-based cancer registries, our use of both registry-derived physician intent to prescribe ET and patient self-reports likely mitigated this concern. This approach is supported by prior research demonstrating a high degree of agreement between self-reported ET use and medical records, with one study reporting strong concordance. (kappa = 0.94) [49].

The timing of certain variables, such as patient trust, also raises the possibility of reverse causality. As surveys were administered after diagnosis and, for many participants, after completion of first-course treatment, reported trust may have been shaped by treatment experiences rather than solely reflecting pre-existing attitudes.

The absence of data on the race or ethnicity of treating physicians limited our ability to explore the potential influence of patient-physician racial concordance. Including this information in future research could provide important insights into how physician characteristics interact with patient experiences and treatment outcomes. [49] A major strength of this study is the utilization of data from the population-based Detroit ROCS cohort, covering patients from the tri-county area of Detroit, Michigan. This minimizes selection bias compared to hospital-based studies, and ensures that our findings are based on a diverse and representative population that enhances the generalizability of the study’s conclusions to a broader population of Black breast cancer patients. Detroit ROCS also conducts yearly follow-up among its participants, allowing us to evaluate use of ET at multiple timepoints and account for changes in ET use over time. Conducting this study among a population of Black cancer patients, who are most likely to experience adverse outcomes, and who remain underrepresented in biomedical research, is also a major strength.

The goal of this exploratory analysis was to enhance understanding of the factors influencing ET prescription and self-reported use among Black breast cancer survivors with HR + disease. The study identified several clinical and patient-level characteristics that were associated with these outcomes. Our study highlights a significant gap between the recommendation and use of ET among these patients and patterns that warrant confirmation and further investigation in future studies. There is evidence indicating that early discontinuation and non-adherence to endocrine treatment is linked to increased mortality among breast cancer patients [11, 50] while ET is known to reduce the risk of breast cancer recurrence and overall mortality [50, 51], these side effects can significantly impact a patient’s quality of life and lead to non-adherence. Therefore, the development of targeted, multilevel interventions that address both clinical and patient-level barriers will be critical to increase guideline-concordant prescription and use of ET. Further research is warranted to develop interventions and ensure they are effectively integrated into standard HR + breast cancer care.

Our study provides valuable insights into the factors influencing both physician intent to prescribe and patient-reported use of endocrine therapy (ET) among Black breast cancer survivors within the Detroit ROCS cohort. Findings show that clinicians are less likely to prescribe ET to patients with lower levels of hormone receptor (HR) positivity, even though guidelines recommend ET for all HR + cases. Patients with lower HR staining (1–10% and 11–90%) or unknown HR staining were significantly less likely to receive an ET recommendation compared to those with higher HR positivity. Additionally, women with regionally staged tumors had a higher likelihood of receiving an ET recommendation or prescription than those with localized tumors.

In terms of patient-reported ET usage, women with weakly HR-positive tumors or unknown HR staining were notably less likely to report using ET compared to those with strongly HR-positive tumors. This discrepancy in ET usage highlights how patient demographics, health literacy, and socioeconomic factors impact adherence to ET, emphasizing the critical role of health literacy in reported treatment continuation.

These findings are significant as they reveal potential disparities in ET treatment practices and adherence among Black breast cancer survivors. Identifying the specific clinical and patient-level factors associated with ET prescription versus self-reported use can guide the development of multilevel interventions to increase ET uptake, ultimately aiming to improve survival outcomes for Black women with HR + breast cancer. Addressing these disparities can help to create a more equitable standard of breast cancer care and support better long-term outcomes in this vulnerable population.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

KSP, LMH, and AMF: conceptualized and designed the study. JJR, AGS, and JLB: performed data collection and curation. SK, KSP, and AMF: conducted statistical analysis and data interpretation. AMF: drafted the manuscript. All authors contributed to manuscript revisions, provided critical feedback, and approved the final version of the manuscript.

Funding

This work was supported by the American Cancer Society (ACS RSG-22–109-01-CTPS), the National Cancer Institute of the National Institutes of Health (U01 CA199240 and CA199240-04S1), the Epidemiology Research Core through a grant awarded to the Karmanos Cancer Institute at Wayne State University (P30CA022453), and the Howard Hughes Medical Institute Gilliam Fellowship.

Data availability

Data sharing is available on approved request by qualified researchers at detroitrocs.org.

Declarations

Conflict of interest

The authors have not disclosed any competing interests.